JP4583093B2 - Bit rate extended speech encoding and decoding apparatus and method - Google Patents

Abstract

Provided are a signal-to-noise ratio (SNR) bitrate scalable speech coding and decoding apparatus and a method of using the same. The coding apparatus includes a base layer (100), a speech quality enhancement layer (130), and a multiplexer (140). The base layer filters an input speech signal using linear prediction coding (104) and generates an excitation signal corresponding to the filtered speech signal through fixed codebook search (117) and adaptive codebook search (115). The speech quality enhancement layer searches a fixed codebook (131) using parameters obtained through the fixed codebook search in the base layer, or searches the fixed codebook using a target signal, which is obtained by removing a contribution of a fixed codebook of the base layer and a signal which is obtained by synthesizing and filtering a previous fixed codebook of the speech quality enhancement layer from a target signal for the fixed codebook search of the base layer. The multiplexer (140) multiplexes signals generated by the base layer and the at least one speech quality enhancement layer. <IMAGE>

Description

  The present invention relates to a speech codec that uses a Code Excited Linear Prediction (CELP) algorithm, and more particularly, speech coding and decoding that extends a signal-to-noise ratio (SNR) bit rate to improve sound quality. The present invention relates to an apparatus and a method thereof.

  A speech codec having a CELP structure is currently most widely used in mobile communication systems, and is based on linear prediction coding (LPC). A voice codec having such a CELP structure differs in required transmission rate and bandwidth depending on the type of service.

  However, since the transmission rate and bandwidth are set by the encoding device in a general audio codec, the transmission rate and bandwidth cannot be selected by the decoding device. Also, when performing multicasting to transmit packet information from one transmitting end to several receiving ends on the network, if the audio codec at the transmitting end has a fixed bit rate, a different bit rate is required. The quality of packet information transmitted to the receiving end can be reduced.

  In order to improve this, a speech codec adopting the bit rate extended speech coding scheme has been proposed. Such an audio codec configures a bitstream so that not only information on the basic codec but also information that makes the signal to be restored more accurate is added.

  Existing bit rate extended speech coding schemes can be broadly divided into SNR bit rate extension methods and bandwidth extension methods.

  In speech coding by the SNR bit rate extension method, a speech signal is encoded by a hierarchical coding method and then decoded. That is, the audio signal is encoded by dividing the audio signal into a basic layer and a sound quality improvement layer. The basic layer transmits only information that can restore the minimum sound quality. In the sound quality improvement layer, additional information that can improve sound quality is transmitted.

  However, the existing proposed SNR bit rate extended speech encoding apparatus is configured to encode the base layer and the sound quality improvement layer independently. Therefore, since the calculation for detecting the correlation between the target signal (or target vector) required when searching the fixed codebook, the impulse response, and the energy is performed in the basic layer and the sound quality improvement layer, respectively. Therefore, a large amount of computation is required to obtain a parametric variable for fixed codebook search.

  In addition, the proposed SNR bit rate extended speech coder has changed the structure of an existing standardized CELP speech coder in order to additionally operate the sound quality enhancement layer. It has the disadvantage that it is not compatible with the CELP speech encoder.

  The technical problem to be solved by the present invention includes a fixed codebook of an existing standardized audio codec and a fixed codebook having a multilayer structure, and an SNR bit rate compatible with an existing standardized audio codec. The present invention provides a speech encoding and decoding apparatus and method for extending the above.

  Another technical problem to be solved by the present invention is to provide an SNR bit rate enhanced speech coding and decoding apparatus and method with a reduced amount of computation for determining a parameter for searching a fixed codebook. .

  Still another technical problem to be solved by the present invention is to use the contribution of the fixed codebook searched in the basic layer and the target signal from which the synthesized excitation signal of the sound quality improvement layer is removed, It is an object of the present invention to provide an SNR bit rate extended speech encoding and decoding apparatus and method for searching a fixed codebook in a sound quality enhancement layer.

  Yet another technical problem to be solved by the present invention is that an algebraic codebook allows a position of a pulse searched in a basic layer and a position of a pulse searched in a sound quality improvement layer to be overlapped. The present invention provides an SNR bit rate extended speech encoding and decoding apparatus and method capable of overcoming these limitations.

  Still another technical problem to be solved by the present invention is to provide an SNR bit rate enhanced speech coding and decoding apparatus and method capable of reducing quantization bits for a gain value of a fixed codebook in a sound quality enhancement layer. By the way.

In speech signal coding apparatus, and a base layer to produce an excitation signal of the linear using predictive coding filter the input audio signal, the filtered audio signal by a fixed codebook search and adaptive codebook search, the The correlation defined by the correlation d (n) between the target signal and the impulse response signal required for the fixed codebook search detected in the basic layer, and the code of the pulse and the correlation d (n) Each pulse, each pulse based on a parametric variable obtained by a fixed codebook search in the basic layer, including a degree of correlation C corresponding to size information of degree d (n) and an energy E of an impulse response signal small code and the speech quality enhancement layer searching fixed codebook by searching the algebraic codebook which associates the position of each pulse, the And a speech signal encoding device including a multiplexer that multiplexes the signal generated in the base layer and the signal generated in the sound quality enhancement layer and outputs the multiplexed signal. To do.

  Here, the fixed codebook search in the basic layer and the fixed codebook search in the sound quality improvement layer are performed using an algebraic codebook.

  Here, the sound quality enhancement layer is a difference between a first gain value obtained by a fixed codebook search performed by the basic layer and a second gain value obtained by a fixed codebook search by the sound quality enhancement layer. It further includes a function to quantize.

  Here, the multiplexer includes quantization information of linear predictive coding coefficients generated in the base layer, a fixed codebook index in the base layer, an adaptive codebook index in the base layer, Quantization information of a fixed codebook gain value and quantization information of an adaptive codebook gain value in the base layer, a fixed codebook index generated in the sound quality enhancement layer, and a fixed codebook gain value of the base layer and the It is characterized in that the quantized information for the difference value between the fixed codebook gain values of the sound quality enhancement layer is multiplexed.

  Here, if there are a plurality of sound quality enhancement layers, the multiplexer multiplexes quantized information related to a difference value between the fixed codebook index output from the plurality of sound quality enhancement layers and the fixed codebook gain value. It is characterized by doing.

Further, in the speech signal encoding apparatus, a base layer that performs linear predictive encoding filtering on an input speech signal and generates an excitation signal corresponding to the filtered speech signal by fixed codebook search and adaptive codebook search; , Defined by the correlation d (n) between the target signal and the impulse response signal necessary for the fixed codebook search detected in the basic layer, the code of the pulse and the correlation d (n), Each pulse includes a correlation degree C corresponding to the size information of the correlation degree d (n) and an energy E of an impulse response signal , based on a parameter generated by a fixed codebook search in the basic layer. , fixed code searching fixed codebook by searching the algebraic codebook which associates code and the position of each pulse of each pulse A book search unit for detecting a difference between a first fixed codebook gain value generated by the fixed codebook search of the base layer and a second fixed codebook gain value output from the fixed codebook search unit; A plurality of sound quality enhancement layers including a gain value difference quantizer for quantizing the difference, and multiplexing the signals generated in the basic layer and the signals generated in the sound quality enhancement layer An audio signal encoding device including an encoder is provided.

Further, in the audio signal decoding apparatus for decoding an audio signal encoded by the encoding device according to any one of the above, the basic layer and at least one sound quality improvement layer are encoded. A first decoding unit for decoding encoded information in the basic layer of the audio signal, and encoded information in the sound quality improving layer of the encoded audio signal according to an operating environment of the audio signal decoding device A second decoding unit that restores the signal, an arithmetic unit that computes the signal restored by the first decoding unit and the signal restored by the second decoding unit according to the operating environment of the audio signal decoding device A speech signal for recovering a speech signal by synthesizing a signal output from the arithmetic unit using a linear predictive coding coefficient output from the first decoding unit; Providing an audio signal decoding apparatus including a source unit, a.

  Here, the first decoding unit includes a linear prediction coding coefficient decoding unit that decodes quantization information of linear prediction coding coefficients included in the coding information in the base layer, and the base layer. A first fixed codebook decoding unit that decodes a fixed codebook index included in the encoded information in the code, and an adaptive codebook that decodes the adaptive codebook index included in the encoded information in the base layer A decoding unit; and a gain value decoding unit that respectively decodes a fixed codebook gain value and an adaptive codebook gain value included in the encoding information in the base layer.

  Here, the second decoding unit includes a gain value difference decoding unit for decoding quantization information of a difference between fixed codebook gain values included in encoding information in the speech enhancement layer; A second fixed codebook decoding unit that decodes a fixed codebook index included in the encoded information in the sound quality enhancement layer.

  Here, the arithmetic unit calculates a difference between the decoded fixed codebook gain value output from the gain value decoding unit and the decoded gain value output from the gain value difference decoding unit. A first adder to be added, and the decoded fixed codebook gain value output from the gain value decoding unit or the gain value output from the first adder according to an operating condition of the audio signal decoding device. A first selection switch to be transmitted, a fixed codebook of a decoded sound quality enhancement layer output from the second fixed codebook decoding unit, and a decoded base layer output from the first fixed codebook decoding unit And a signal output from the second adder or the first fixed codebook depending on operating conditions of the audio signal decoding apparatus. A second selection switch for transmitting the decoded fixed codebook output from a codebook decoding unit; a decoded adaptive codebook output from the adaptive codebook decoding unit; and a gain value decoding unit A first multiplier that multiplies a gain value of the decoded adaptive codebook that is output; a second multiplication that multiplies the signal output from the first selection switch and the signal output from the second selection switch; And a third adder for adding the signal output from the first multiplication branch and the signal output from the second multiplication branch.

  Here, the speech signal restoration unit outputs a synthesis filter that synthesizes a signal output from the third adder using the linear prediction coding coefficient, and is output from the linear prediction coding coefficient and the synthesis filter. And a post-processing unit for obtaining the reconstructed audio signal using a signal.

  Here, the second decoding unit includes a gain value difference decoding unit for decoding quantization information of a difference between fixed codebook gain values included in encoding information in the speech enhancement layer; A fixed codebook decoding unit that decodes a fixed codebook index included in the encoding information in the sound quality enhancement layer.

Extracting a linear predictive coding coefficient of the speech signal input in the base layer and generating an excitation signal corresponding to the input speech signal by fixed codebook search and adaptive codebook search; and at least one In the sound quality enhancement layer, the correlation d (n) between the target signal and the impulse response signal required for the fixed codebook search detected in the basic layer, the code of the pulse, and the correlation d (n) A parameter C defined by the size information of the correlation d (n) and an energy E of an impulse response signal defined in the basic variable and generated by searching the fixed codebook based on each pulse, the steps of searching fixed codebook by searching the algebraic codebook which associates code and the position of each pulse of each pulse, before Providing an audio signal encoding method comprising the steps of multiplexing the signals generated by said speech quality enhancement layer and base layer, an.

  Here, the sound quality improvement hierarchy processing step is performed in a plurality of steps.

  Here, the sound quality enhancement layer processing step includes a difference between the fixed codebook gain value obtained by the fixed codebook search in the basic layer processing step and the gain value obtained by the fixed codebook search of the sound quality improvement layer. It is characterized by quantizing the value.

  Here, the parameter includes the first correlation between the target signal generated in the base layer processing step and the impulse response, the second correlation corresponding to the magnitude of the first correlation, and the energy of the impulse response signal. It is characterized by including.

Here, in the audio signal decoding method for decoding the audio signal encoded into the base layer and at least one sound quality improvement layer by any of the audio signal encoding methods described above, the encoding is performed. Decoding the voice signal, and selectively transmitting the codebook for the base layer and the codebook for the sound quality enhancement layer decoded in the decoding step according to the operation condition of the voice signal encoding, Synthesizing the selectively transmitted codebook and the linear prediction coefficient decoded in the decoding step to generate a reconstructed speech signal.
Here, in the decoding step, the encoded audio signal is demultiplexed into encoded information related to a base layer and encoded information related to a sound quality improvement layer, and the demultiplexed encoded information is decoded. It is characterized by.
Here, the audio signal decoding method adds a difference between a fixed codebook gain value in the base layer decoded in the decoding step and a fixed codebook gain value included in the decoded sound quality enhancement layer. The method further comprises the step of restoring the gain value of the fixed codebook in the sound quality enhancement layer.

  According to the present invention, by presenting a structure in which the bit rate can be expanded without changing an existing standardized CLEP speech coding structure, it is compatible with a system equipped with an existing standardized CLEP speech coding apparatus. It is.

  In addition, according to the above-described embodiment of the present invention, the code searched for in the sound quality improvement layer is obtained by making the fixed codebook search target signal in the base layer similar to the fixed codebook search target signal in the sound quality improvement layer. The book is not saved for the next frame, so it does not affect the operation of the base hierarchy.

  Then, the amount of calculation required for the fixed codebook search in the sound quality improvement layer can be reduced by using the parameter value obtained in the fixed codebook search in the basic layer during the fixed codebook search in the sound quality improvement layer.

  Further, according to another embodiment of the present invention, the target signal required for the fixed codebook search of the sound quality improvement layer, the fixed codebook contribution of the basic layer from the fixed codebook target signal of the basic layer, More accurate by performing a fixed codebook search using the target signal dedicated to the sound quality improvement layer by removing the synthesized signal of the fixed codebook of the previous sound quality improvement layer provided through the synthesis filter of the sound quality improvement layer Can be expected.

  Furthermore, the position of the pulse searched for in the sound quality enhancement layer and the position of the pulse searched for in the basic layer become the same, overcoming the limit point where the pulses of the algebraic codebook have the same magnitude, and the final fixed codebook Therefore, the sound quality of the restored audio signal can be improved.

  Then, the gain value of the sound quality improvement layer is a value obtained by quantizing the difference between the quantized gain value of the base layer and the gain value of the sound quality improvement layer and quantizing the gain value difference having a relatively small dynamic range. By doing so, it is possible to save bits necessary for gain value quantization in the sound quality enhancement layer.

  Hereinafter, the present invention will be described in detail by describing preferred embodiments of the present invention with reference to the drawings. The same reference numerals provided in each drawing denote the same members.

  FIG. 1 is a functional block diagram of a bit rate extended speech encoding apparatus according to a preferred embodiment of the present invention. Referring to FIG. 1, the bit rate extended speech encoding apparatus has a multi-layer fixed codebook structure including a basic layer 100 and a sound quality improvement layer 130.

  In the basic layer 100, encoded information that can restore the minimum sound quality is generated. The base layer 100 is similar to the configuration of existing standardized CELP speech encoders. Therefore, the base layer 100 generates an excitation signal corresponding to the input speech signal by filtering the input speech signal by linear predictive coding.

  The basic layer 100 includes a preprocessing unit 102, an LPC coefficient extraction and vector quantizer 104, a synthesis filter 106, a subtractor 108, a cognitive weighting filter 110, a pitch analysis unit 112, a pitch contribution removal unit 115, and a fixed codebook search. A unit 117, a fixed codebook 119, a first multiplier 121, an adder 123, an adaptive codebook 124, a second multiplier 126, and a gain value quantizer 129.

The preprocessing unit 102 removes the DC component from the audio signal input through the line 101. That is, the preprocessing unit 102 filters the input audio signal using a high-pass filter and removes noise components in the low frequency band of the input audio signal. The high-pass filter H _h1 (n) used has a transfer function as shown in Equation (2).

前処理ユニット１０２から出力される信号は、ライン１０３を通じてＬＰＣ係数の抽出及びベクトル量子化器１０４に伝送される。

The signal output from the preprocessing unit 102 is transmitted to the LPC coefficient extraction and vector quantizer 104 through a line 103.

  The LPC coefficient extraction and vector quantizer 104 extracts the LPC coefficient of the signal output from the preprocessing unit 102. The extracted LPC coefficients are LPC coefficient extraction and vector quantized by the vector quantizer 104. The vector quantization information of the LPC coefficient is transmitted to the synthesis filter 106 and the multiplexer 140 through the line 105.

  The synthesis filter 106 outputs a synthesized signal corresponding to the excitation signal inputted through the line 128 using the vector quantization information of the LPC coefficient. The synthesized signal is output to the subtracter 108 through the line 107.

  The subtracter 108 subtracts the synthesized signal input through the line 107 from the signal output from the preprocessing unit 102 input through the line 103 to generate a difference signal. The difference signal is transmitted to the perceptual weighting filter 110 via line 109.

  The cognitive weighting filter 110 keeps the quantization noise below the masking critical value in order to use the masking effect of the human auditory structure. Accordingly, the cognitive weighting filter 110 outputs a signal including a weight value to the pitch analysis unit 112 so that the quantization noise of the difference signal is minimized.

  The pitch analysis unit 112 searches for an open circuit pitch and a closed circuit pitch for the signal output from the cognitive weighting filter 110. That is, the pitch analysis unit 112 divides the signal output from the cognitive weighting filter 110 into a plurality of subframes, analyzes the pitch of each subframe, and outputs the index and gain value of the adaptive codebook. The index of the adaptive codebook is transmitted to the multiplexer 140 through the line 114 while being transmitted to the pitch contribution removing unit 115 and the adaptive codebook 124 through the line 113. The gain value of the adaptive codebook is provided to a gain value quantizer 129.

The pitch contribution removing unit 115 detects a target signal (or target vector) necessary for fixed codebook search from the output signal of the cognitive weighting filter 110 based on the index of the adaptive codebook 124. Then, the pitch contribution removal unit 115 subtracts the pitch contribution y ₁ (n) from the line 111 to obtain the fixed codebook search target signal, and the line 116 to the fixed codebook search unit 117 of the base layer 100 to improve the sound quality. The data is output to the fixed codebook search unit 131 in the hierarchy 130. The pitch contribution y ₁ (n) is obtained by Expression (3).

数式（３）でＡＣ_Ｇ（ｎ）は適応コードブック利得値が乗算された値である。

In Equation (3), AC _G (n) is a value obtained by multiplying the adaptive codebook gain value.

  The fixed codebook search unit 117 uses the target signal x ′ (n) input through the line 111 to determine the degree of correlation d (n) between the target signal and the impulse response h (n).

  For example, assuming that the size of the subframe is 40 samples and the number of pulses in each layer is four, the correlation d (n) can be defined as Equation (4).

数式（４）で、ｈ（ｉ−ｎ）はインパルス応答であり、ｘ’（ｎ）は対象信号である。

In Expression (4), h (i−n) is an impulse response, and x ′ (n) is a target signal.

  The impulse response h (n) and the correlation d (n) are provided to the fixed codebook search unit 131 of the sound quality enhancement layer 130 through a line 118 '.

  Based on the impulse response h (n) and the correlation d (n), the fixed codebook search unit 117 searches for a fixed codebook having an algebraic codebook structure as shown in Table 1. Explore.

表１を参考すれば、固定コードブック探索部１１７で固定コードブックベクトルは４個の位置でのみそのパルスのサイズが０でない。したがって、前記パルスの符号ｓと相関度ｄ（ｎ）とを利用して相関度Ｃは数式（５）のように定義されうる。固定コードブック探索部１１７は、数式（５）により相関度Ｃを検出する。

Referring to Table 1, the fixed codebook search unit 117 has a fixed codebook vector whose pulse size is not 0 only at four positions. Accordingly, the correlation degree C can be defined as Equation (5) using the pulse code s and the correlation degree d (n). Fixed codebook search unit 117 detects correlation degree C using equation (5).

数式（５）で、ｍ_ｉはｉ番目パルスの位置を表し、ｓ_ｉはｉ番目パルスの符号を表す。

In Equation (5), _{m i} denotes the position of i-th pulse, _{s i} represents the sign of the i-th pulse.

  The fixed codebook detection unit 117 detects the energy E of the impulse response h (n) of the synthesis filter 106 using Equation (6).

数式（６）で、Φ（ｍ_ｉ、ｍ_ｊ）はｉ番目パルスの位置及びｊ番目パルスの位置に対するインパルス応答信号ｈ（ｎ）間の相関度であり、ｓ_ｉはｉ番目パルスの符号であり、ｓ_ｊはｊ番目パルスの符号である。

In Equation (6), Φ (m _i , m _j ) is the degree of correlation between the position of the i th pulse and the impulse response signal h (n) with respect to the position of the j th pulse, and s _i is the sign of the i th pulse. Yes, s _j is the sign of the jth pulse.

  The fixed codebook search unit 117 stores the correlation C and the energy E of the impulse response h (n). Correlation degree C is divided and stored in code sign [d (i)] and its absolute value. sign [d (i)] is the sign of d (i). The energy E is stored in a form such as Equation (7).

エネルギーＥに対する数式（６）は、数式（８）のように再定義されうる。

Equation (6) for energy E can be redefined as equation (8).

固定コードブック探索部１１７は、前記検出された相関度ＣとエネルギーＥとをライン１１８”を通じて、音質向上階層１３０の固定コードブック探索部１３１に提供しながら、検出された相関度ＣとエネルギーＥとを利用して固定コードブックを探索する。前記固定コードブック探索により、固定コードブックインデックスと利得値とが得られれば、固定コードブック探索部１１７は、前記固定コードブックインデックスを固定コードブック１１９と多重化器１４０とに伝送し、前記利得値を利得値量子化器１２９に伝送する。

The fixed codebook search unit 117 provides the detected correlation C and energy E to the fixed codebook search unit 131 of the sound quality enhancement layer 130 through the line 118 ″ while providing the detected correlation C and energy E. If a fixed codebook index and a gain value are obtained by the fixed codebook search, the fixed codebook search unit 117 searches the fixed codebook 119 for the fixed codebook index. Are transmitted to the multiplexer 140, and the gain value is transmitted to the gain value quantizer 129.

  The fixed codebook 119 outputs a fixed codebook vector of the base layer 100 based on the index input through the line 118. The fixed codebook vector output from the fixed codebook 119 is provided to the first multiplier 121 through the line 120.

The first multiplier 121, a quantized gain value G _c is multiplied by the fixed codebook vector for said gain value of the fixed codebook provided in gain value quantizer 129, and outputs the result via a line 122. The signal output through line 122 is a fixed codebook vector. The quantization gain value G _c is provided from a gain value quantizer 129.

  If the adaptive codebook index is applied through the line 113, the adaptive codebook 124 outputs pulse position information and code information corresponding to the adaptive codebook index. The adaptive codebook vector output through line 125 is provided to the second multiplier 126.

The second multiplier 126, the adaptive code gain value G _p quantized relative gain values of the book by multiplying the adaptive codebook vector transmitted via the line 125, and outputs the result via a line 127. Signal output through the line 127 is a vector of adaptive codebook gain value G _p is multiplied. The quantized gain value G _p is provided from a gain value quantizer 129.

The adder 123, the adaptive signal obtained by multiplying the fixed codebook vector with quantized gain value G _c is input via a line 122, a quantized gain value G _p is input via the line 127 An excitation signal is obtained by adding the signal obtained by multiplying the codebook vector. The excitation signal is output to the synthesis filter 106 through the line 128.

The gain value quantizer 129 quantizes the gain value of the fixed codebook output from the fixed codebook search unit 117 and the gain value of the adaptive codebook output from the pitch analysis unit 112. Gain value G _c of the gain value obtained by quantizing the fixed codebook is output to the first multiplier 121, a gain value G _p the gain value obtained by quantizing the adaptive codebook is output to the second multiplier 126 The The quantized gain value G _c is also provided to a gain value difference quantizer 134 included in the sound quality enhancement layer 130.

  The sound quality enhancement layer 130 is for providing additional bits in addition to the bits provided in the basic layer 100 in order to improve the restored sound quality. For example, when the base layer 100 provides a bit rate of 8 kbps, the sound quality enhancement layer 130 can provide an additional bit rate of 4 kbps. Although FIG. 1 illustrates a configuration in which one voice enhancement layer 130 is connected to the basic layer 100 for convenience of explanation, a plurality of voice enhancement layers may be connected to the basic layer 100.

  The sound quality enhancement layer 130 includes a fixed codebook search unit 131 and a gain value difference quantizer 134. The fixed codebook search unit 131 includes an impulse response signal h (n) provided through the line 118 ′, a correlation degree d (n) between the target signal and the impulse response signal h (n), a pulse code, and the correlation. The fixed codebook is searched using the degree of correlation C and the energy E of the impulse response signal h (n) corresponding to the size information of d (n) detected using the degree d (n). .

In this way, the fixed codebook search unit 131 performs a fixed codebook search for the same target signal as the target signal searched by the fixed codebook search unit 117. The fixed codebook search unit 131 uses an algebraic codebook. The fixed codebook search unit 131 finds a vector _ck that minimizes the MSE (Mean Square Error) of the target signal (target vector) and maximizes Equation (9). The found vector _ck becomes a fixed codebook vector.

数式（９）でΦはインパルス応答ｈ（ｎ）間の相関度を表す。前記ｄ（ｎ）とΦとは基本階層１００で提供する値を利用する。前記Φは、固定コードブック探索部１１７から提供される。したがって、固定コードブック探索部１３１は固定コードブック探索時に必要な演算量を減らせる。

In Equation (9), Φ represents the degree of correlation between impulse responses h (n). As the d (n) and Φ, values provided in the basic layer 100 are used. The Φ is provided from the fixed codebook search unit 117. Therefore, the fixed codebook search unit 131 can reduce the amount of calculation required for the fixed codebook search.

  Assuming that the order of the fixed codebook vector of the basic layer 100 is 40, and that four non-zero magnitude pulses are to be searched for in the basic layer 100 and the sound quality enhancement layer 130, respectively, the fixed codebook 117 of the basic layer 100 starts with Four pulses are searched, and the fixed codebook search unit 131 of the sound quality improvement layer 130 searches for four pulses. Therefore, the fixed codebook search unit 131 also considers the influence of the four pulses searched for in the basic hierarchy 100. Therefore, the correlation degree C ′ obtained by the fixed codebook search unit 131 can be defined as Equation (10), and the energy E ′ can be defined as Equation (11).

数式（５）に定義された相関度Ｃ値を利用して、前記数式（１０）は、数式（１２）のように再定義されうる。

Using the correlation degree C value defined in Equation (5), Equation (10) can be redefined as Equation (12).

固定コードブック探索部１３１は、探索過程の複雑度を減らすためにエネルギーＥ’を数式（１３）のように再定義された演算により検出できる。

The fixed codebook search unit 131 can detect the energy E ′ by a redefined calculation as in Equation (13) in order to reduce the complexity of the search process.

数式（１３）は、数式（８）に定義されているエネルギーＥを利用すれば、数式（１４）のように再定義されうる。

Equation (13) can be redefined as Equation (14) by using the energy E defined in Equation (8).

相関度Ｃ’とエネルギーＥ’とは、音質向上階層１３０での固定コードブック探索以前に保存されて、固定コードブック探索過程を簡素化できる。

The correlation degree C ′ and the energy E ′ are stored before the fixed codebook search in the sound quality enhancement layer 130, and the fixed codebook search process can be simplified.

  The process of the fixed codebook search unit 131 for obtaining the code information and the position information of the pulse of the sound quality enhancement layer 130 using the correlation degree C ′ and the energy E ′ described above is the fixed codebook search of the basic layer 100. This is performed in the same manner as that performed by the unit 117. At this time, the pulse position information searched in the basic layer 100 and the pulse position information searched in the sound quality improvement layer may be the same.

  FIG. 2 is a diagram for explaining the pulse positions searched by the fixed codebook search unit 117 and the pulse positions searched by the fixed codebook search unit 131 in the bit rate extended speech encoding apparatus of FIG. It is.

  Referring to FIG. 2, the position of the pulse searched in the fixed codebook search 201 may be the same as the position of the pulse searched in the sound quality improvement hierarchical fixed codebook search 202. Therefore, the pulse size of the final fixed codebook has a multiplexed size (hereinafter referred to as a multiplexed size) including the sizes of the fixed codebook pulses of the basic layer 100 and the sound quality enhancement layer 130. Therefore, the pulse size of the algebraic codebook is not only +1 or -1.

  The fixed codebook search unit 131 provides the fixed codebook vector obtained from the search result to the multiplexer 140, and provides the gain value of the fixed codebook to the gain value difference quantizer 134. The fixed codebook index in the sound quality enhancement layer 130 may be composed of pulse code information and pulse position information.

  Since the fixed codebook index searched in the sound quality enhancement layer 130 is not stored for the next frame, the operation of the basic layer 100 is not affected.

Quantizer 134 of the gain value difference, determining a difference between the gain value G _c of the quantized fixed codebook gain value 132 and base layer 100 of the fixed codebook obtained by the fixed codebook searching unit 131, the Quantize the difference. As a result, the gain value difference quantization information G _diff is transmitted from the gain value difference quantizer 134 to the multiplexer 140 through the line 135, so that the sound quality enhancement layer 130 quantizes the gain value of the fixed codebook. Bits can be reduced.

  The multiplexer 140 is a fixed code of the sound quality improvement layer provided from the LPC coefficient quantization information provided from the base layer 100, a fixed codebook index, an adaptive codebook index, gain value quantization information, and the sound quality improvement layer 130. The quantization information of the book index and gain value difference is output to the bit stream.

  The bit streams of the basic layer 100 and the sound quality improvement layer 130 are transmitted separately. That is, as shown in FIG. 1, the bit stream of the sound quality enhancement layer 130 is transmitted after the bit stream of the base layer 100. Accordingly, the bit stream can be easily separated at a bit rate required for the decoding apparatus according to network traffic conditions. For example, when the channel characteristics on the decoding device side are poor and only the base layer bit stream can be received, the decoding device uses the bits of the base layer among the bit streams transmitted by the bit rate extended speech encoding device of FIG. Only the stream can be received.

  FIG. 3 is a block diagram of a bit rate extended speech decoding apparatus according to an embodiment of the present invention.

  Referring to FIG. 3, the bit rate extended speech decoding apparatus includes a demultiplexer 301, an LPC coefficient decoding unit 302, a gain value decoding unit 303, a first fixed codebook decoding unit 304, an adaptive codebook. Decoding unit 305, gain value difference decoding unit 306, second fixed codebook decoding unit 307, first adder 308, second adder 309, first selection switch 310, second selection switch 311, first It includes a multiplier 312, a second multiplier 313, a third adder 314, a synthesis filter 315, and a post-processing unit 316.

  The bit rate extended speech decoding apparatus can selectively receive a bit stream transmitted from the bit rate extended speech encoding apparatus. That is, if only the bitstream for the base layer is received from the bitstream, the sound quality of the base layer can be restored, and if both the bitstreams for the base layer and the sound quality improvement layer are received, further improved sound quality can be provided.

  The demultiplexer 301 demultiplexes the received bit stream into the information of each module and outputs it. That is, the demultiplexer 301 decodes LPC coefficient quantization information to the LPC coefficient decoding unit 302, gain value quantization information to the gain value decoding unit 303, and gain value difference quantization information to decode the gain value difference. In the second codec decoding unit 307, the fixed codebook index in the first fixed codebook decoding unit 304, and the adaptive codebook index in the adaptive codebook decoding. Provided to the conversion unit 305.

  The structure of the LPC coefficient decoding unit 302 is determined by the LPC coefficient extraction and vector quantizer 104 on the encoding device side, and restores the LPC coefficient from the input LPC coefficient quantization information. The restored LPC coefficient is provided to the synthesis filter 315 and the post-processing unit 316.

The structure of gain value decoding section 303 is determined by gain value quantizer 129 on the encoding device side. The gain value decoding unit 303 decodes input gain value quantization information. The gain value quantization information includes an adaptive codebook gain value and a fixed codebook gain value. Thus, the adaptive codebook gain value g _p at the base layer 100 and the fixed codebook gain value g _c from the gain value decoding unit 303 are output.

  The first fixed codebook decoding unit 304 decodes the input fixed codebook index of the basic layer 100 and outputs the fixed codebook of the basic layer 100. The fixed codebook decoding method is determined by the search method in fixed codebook search unit 117 of the encoding device. The adaptive codebook decoding unit 305 decodes the input adaptive codebook index and outputs the adaptive codebook of the base layer 100.

  The PC coefficient decoding unit 302, the gain value decoding unit 303, the first fixed codebook decoding unit 304, and the adaptive codebook decoding unit 305 are the codes in the base layer 100 transmitted from the demultiplexer 301. It may be defined as a first decoding unit that decodes the encoded information.

  The operations of the gain value difference decoding unit 306 and the second fixed codebook decoding unit 307 depend on the network traffic state and the processing capacity of the receiving terminal.

  If it is determined that the gain value difference decoding unit 306 and the second fixed codebook decoding unit 307 are operated, the gain value difference decoding unit 306 may input the gain value difference quantization information. Decode. The second fixed codebook decoding unit 307 decodes the input fixed codebook index of the sound quality improvement layer. The gain value difference decoding method is determined by the gain value difference quantizer 134 on the encoding device side. The decoding method in the second fixed codebook decoding unit 307 is determined by the second fixed codebook search unit 131 on the encoding device side.

  The gain value difference decoding unit 306 and the second fixed codebook decoding unit 307 are regarded as a second decoding unit for decoding the encoded information in the sound quality enhancement layer 130 transmitted from the demultiplexer 301. sell.

The first adder 308 includes a decoded fixed codebook gain value g _c output from the gain value decoding unit 303 and a decoded gain value difference g _diff output from the gain value difference decoding unit 306. Is added. The output of the first adder 308 is the gain value of the sound quality enhancement layer at the time of decoding.

  The second adder 309 includes a fixed codebook of the sound quality enhancement layer 130 decoded by the second fixed codebook decoding unit 307 and a fixed codebook of the basic layer 100 decoded by the first fixed codebook decoding unit 304. And add. Therefore, the signal output from the second adder 309 can be defined as Equation (15).

数式（１５）で、ｃ（ｎ）は基本階層での固定コードブックであり、ｃ'（ｎ）は音質向上階層での固定コードブックである。

In Equation (15), c (n) is a fixed codebook in the basic layer, and c ′ (n) is a fixed codebook in the sound quality improvement layer.

  Accordingly, the fixed codebook pulse in the decoding apparatus has an algebraic codebook pulse structure having a multiple size by accumulating the algebraic codebooks of the basic layer and the sound quality enhancement layer. Accumulating the algebraic codebook is intended to compensate for the disadvantages that occur in existing fixed codebook structures where every pulse has the same magnitude. Therefore, the accumulated pulse of the algebraic codebook has a code suitable for the target signal.

The first selection switch 310 selectively transmits the fixed codebook gain value g _c decoded by the gain value decoding unit 303 and the signal output from the first adder 308. That is, if the decoding apparatus operates in the base layer, the first selection switch 310 transmits the fixed codebook gain value g _c output from the gain value decoding unit 303, and the corresponding decoding apparatus performs the sound quality improvement layer. The first selection switch 310 transmits the gain value output from the adder 308.

  The second selection switch 311 selectively transmits the signal output from the second adder 309 and the fixed codebook in the base layer 100 output from the first fixed codebook decoding unit 304. That is, when the decoding device does not operate in the sound quality enhancement layer, the second selection switch 311 transmits the signal output from the first fixed codebook decoding unit 304, and the decoding device is in the sound quality enhancement layer. In operation, the second selection switch 311 transmits a signal output from the second adder 309.

  The first multiplier 312 multiplies the fixed codebook output from the second selection switch 311 by the gain value output from the first selection switch 310 and outputs the result.

The second multiplier 313, and outputs the multiplied gain value g _p adaptive codebook output from the gain value decoding unit 303 to the decoded adaptive codebook output from the adaptive codebook decoding unit 305.

  The third adder 314 adds the information related to the fixed codebook output from the first multiplier 312 and the information related to the adaptive codebook output from the second multiplier 313 to generate a restored excitation signal.

  The first adder 308, the second adder 309, the third adder 314, the first multiplier 312, the second multiplier 313, the first selection switch 310, and the second selection switch 311 described above are the first decoding described above. The decoding unit and the second decoding unit may be defined as a calculation unit that calculates the signals decoded by the decoding device according to the operating environment of the decoding device.

  The synthesis filter 315 uses the restored LPC coefficient provided from the LPC coefficient decoding unit 302 to synthesize the excitation signal provided from the adder 314 and restore the speech signal.

  The post-processing unit 316 improves the sound quality of the audio signal transmitted from the synthesis filter 315. That is, in order to improve the sound quality of the audio signal, the post-processing unit 316 uses the LPC coefficient provided from the LPC coefficient decoding unit 302 to filter the signal output from the synthesis filter 315. Is used.

  The synthesis filter 315 and the post-processing unit 316 described above may be defined as a restoration unit that synthesizes the signal output from the arithmetic unit with the LPC coefficient output from the LPC coefficient decoding unit 302 to restore the speech signal.

  FIG. 4 is a flowchart of a bit rate extended speech encoding method according to an embodiment of the present invention.

  In operation 401, the audio signal encoding apparatus preprocesses an input audio signal as in the preprocessing unit 102 of FIG. In operation 402, the speech signal encoding apparatus extracts LPC coefficients from the preprocessed speech signal, and generates quantization information of the extracted LPC coefficients.

  In operation 403, the speech signal encoding apparatus uses the generated LPC coefficient quantization information to synthesize the excitation signal as in the synthesis filter 106 of FIG. In operation 404, the speech signal encoding apparatus detects an LPC residual signal by subtracting the synthesized signal from the preprocessed signal. In operation 405, the speech signal encoding apparatus filters the detected LPC residual signal as in the perceptual weighting filter 110 of FIG. 1 and outputs a perceptual weighted signal.

  In operation 406, the speech signal encoding apparatus analyzes the perceptually weighted signal pitch as in the pitch analysis unit 112 of FIG. 1 to obtain an index and gain value of the adaptive codebook. Then, the speech signal encoding apparatus removes the pitch contribution from the signal weighted based on the index of the adaptive codebook as in the pitch contribution removal unit 115 in FIG. The target signal necessary for this is detected.

  In operation 407, the speech signal encoding apparatus searches the base layer fixed codebook as in the first fixed codebook search unit 117 of FIG. 1 to obtain a fixed codebook gain value and a fixed codebook index. Generate. In operation 408, the speech signal encoding apparatus quantizes the detected fixed codebook gain value and the detected adaptive codebook gain value as in the gain value quantizer 129 of FIG. .

  In operation 409, the speech signal encoding apparatus searches for a sound quality enhancement layer fixed codebook using parameters such as the correlation degree C and d (n) and energy E in the base layer. Through the sound quality improvement layer fixed codebook search, the gain value and the index of the fixed codebook of the sound quality improvement layer are respectively generated.

  In operation 410, the speech signal encoding apparatus quantizes the difference between the gain value of the base layer fixed codebook and the gain value of the fixed codebook of the sound quality enhancement layer. The fixed codebook search and gain value quantization process in the sound quality enhancement layer may be performed in a plurality of layers as described with reference to FIG. If the processing of the sound quality improvement layer is divided into a plurality of layers, the quality of the restored audio signal can be improved.

  In operation 411, the speech signal encoding apparatus obtains the LPC coefficient quantization information obtained through the above-described steps, the base layer fixed codebook index, the base layer adaptive codebook index, the base layer fixed codebook gain value, The base layer adaptive codebook gain value, the sound quality enhancement layer fixed codebook index, and the gain value difference quantization information are multiplexed into a bitstream format and transmitted to the audio signal decoding apparatus side.

  FIG. 5 is a flowchart of a bit rate enhanced speech decoding method according to an embodiment of the present invention.

  In operation 501, the audio signal decoding apparatus demultiplexes the received bitstream into information of each component like the demultiplexer 301 in FIG. 3.

  In operation 502, the audio signal decoding apparatus decodes the demultiplexed signal. That is, LPC coefficient decoding section 302, gain value decoding section 303, first fixed codebook decoding section 304, adaptive codebook decoding section 305, gain value difference decoding section 306, second fixed code in FIG. The demultiplexed signal is decoded as in the book decoding unit 307.

  In operation 503, the speech signal decoding apparatus restores the sound quality enhancement layer fixed codebook gain value by a predetermined calculation process. The speech signal decoding apparatus adds the decoded fixed codebook gain value and the gain value difference received as quantization information of the fixed codebook gain value of the sound quality enhancement layer, and adds the fixed code of the sound quality enhancement layer. Restore the book gain value.

  In operation 504, the audio signal decoding apparatus selectively transmits the fixed codebook of the sound quality enhancement layer and the fixed codebook of the basic layer according to the operating conditions of the audio signal decoding apparatus, and also selectively transmits the gain value. Is done. That is, if the speech signal decoding apparatus operates in the sound quality enhancement layer, the fixed codebook in the sound quality enhancement layer multiplied by the gain value of the restored fixed codebook in the sound quality enhancement layer is transmitted. On the other hand, if the audio signal encoding apparatus does not operate in the sound quality enhancement layer, the fixed codebook obtained by multiplying the decoded code for the basic layer by the gain value of the fixed codebook for the basic layer is transmitted.

  In operation 505, the speech signal decoding apparatus synthesizes the fixed codebook selectively transmitted in operation 504 using the LPC coefficients decoded in operation 502.

  In operation 506, the speech signal decoding apparatus generates a speech signal restored by post-processing as in the post-processing unit 316.

  FIG. 6 is a functional block diagram of a bit rate extended speech encoding apparatus according to another embodiment of the present invention. Referring to FIG. 6, the bit rate extended speech encoding apparatus has a multi-layer fixed codebook structure including a basic layer 600 and a sound quality improvement layer 630.

  In the basic layer 600, encoded information that can restore the minimum sound quality is generated. Base layer 600 is similar to the configuration of existing standardized CELP speech encoders. Accordingly, the base layer 600 filters the input speech signal by linear predictive coding, and generates an excitation signal corresponding to the filtered speech signal. The excitation signal is generated by a fixed codebook search and an adaptive codebook search.

  The basic layer 600 includes a preprocessing unit 602, LPC coefficient extraction and vector quantizer 604, synthesis filter 606, subtractor 608, cognitive weighting filter 610, pitch analysis unit 612, pitch contribution removal unit 615, fixed codebook. A search unit 617, a fixed codebook 619, a first multiplier 621, an adder 623, an adaptive codebook 624, a second multiplier 626, and a gain value quantizer 629 are configured.

  The preprocessing unit 602 removes the DC component from the audio signal input through the line 601. That is, the preprocessing unit 602 filters the input audio signal using a high-pass filter to remove a low frequency band noise component of the input audio signal. The high-pass filter used is the same as the high-pass filter of the preprocessing unit 102 of the basic hierarchy 100 in one embodiment of the present invention. The signal output from the preprocessing unit 602 is transmitted to the LPC coefficient extraction and vector quantizer 604 through a line 603.

  The LPC coefficient extraction and vector quantizer 604 extracts the LPC coefficient of the signal output from the preprocessing unit 602. The extracted LPC coefficient is vector quantized by the LPC coefficient extraction and vector quantizer 604. The vector quantization information of the LPC coefficient is transmitted to the synthesis filter 606 and the multiplexer 650 through the line 605.

  The synthesis filter 606 outputs a synthesized signal corresponding to the excitation signal inputted through the line 628 using the vector quantization information of the LPC coefficient. The synthesized signal is output to a subtracter 608 through a line 607.

  The subtracter 608 generates an LPC residual signal by subtracting the synthesized signal input through the line 607 from the signal output from the preprocessing unit 602 input through the line 603. The LPC residual signal is transmitted to the cognitive weighting filter 610 through line 609.

  The cognitive weighting filter 610 keeps the quantization noise below the masking critical value in order to use the masking effect of the human auditory structure. Accordingly, the cognitive weighting filter 610 outputs a signal including a weight value to the pitch analysis unit 612 so that the quantization noise of the LPC residual signal is minimized.

  The pitch analysis unit 612 searches for an open circuit pitch and a closed circuit pitch for the signal output from the cognitive weighting filter 610. That is, the pitch analysis unit 612 divides the signal output from the cognitive weighting filter 610 into a plurality of subframes, analyzes the pitch of each subframe as in the standardized CLEP speech coding apparatus, and performs an adaptive codebook. Output the index and gain value.

  The index of the adaptive codebook is transmitted to the multiplexer 650 through the line 614 while being transmitted to the pitch contribution removing unit 615 and the adaptive codebook 624 through the line 613. Also, the gain value of the adaptive codebook is provided to a gain value quantizer 629.

The pitch contribution removal unit 615 outputs a target signal necessary for a fixed codebook search from the output signal of the cognitive weighting filter 610 based on the index of the adaptive codebook. Then, the pitch contribution removal unit 615 subtracts the pitch contribution y ₁ (n) from the output signal of the cognitive weighting filter 610 to obtain the target signal necessary for the fixed codebook search through the line 616. The result is output to a fixed codebook search unit 617 of 600. The pitch contribution y ₁ (n) is obtained by Expression (3).

  The fixed codebook search unit 617 obtains the degree of correlation d (n) between the target signal and the impulse response h (n) using the target signal x ′ (n) input through the line 611.

  For example, assuming that the size of the subframe is 40 samples and the number of pulses in each layer is four, the correlation d (n) can be defined as Equation (2).

  The fixed codebook search unit 617 searches for a fixed codebook in the form of an algebraic codebook configured as in the example of Table 1 based on the impulse response h (n) and the correlation degree d (n). To do. Referring to Table 1, the pulse size of the fixed codebook vector in the fixed codebook search unit 617 is not 0 only at the four positions. Therefore, the correlation degree C corresponding to the magnitude of the correlation degree d (n) using the pulse code s and the correlation degree d (n) can be defined as Equation (3). Fixed codebook search unit 617 detects correlation degree C using equation (3). The fixed code book detection unit 617 detects the impulse response energy E by the mathematical formula (4).

  The fixed codebook search unit 617 stores the correlation C and energy E. Fixed codebook search unit 617 stores correlation degree C by dividing it into code sign [d (i)] and its absolute value. sign [d (i)] is the sign of d (i). The energy E is stored in a form such as Equation (5). Equation (4) for energy E can be redefined as equation (6).

  If a fixed codebook index and a gain value are obtained by the search, the fixed codebook search unit 617 transmits the fixed codebook index to the fixed codebook 619 and the multiplexer 650, and the gain value is gained. Transmit to the value quantizer 629.

  Fixed codebook 619 outputs a fixed codebook vector of base layer 600 based on the index input through line 618. The fixed codebook vector is configured based on the pulse position information m and the code information s. The fixed codebook vector output from the fixed codebook 619 is provided to the first multiplier 621 through a line 620.

The first multiplier 621 multiplies the fixed codebook vector by the quantized gain value G _c for the gain value of the fixed codebook provided from the gain value quantizer 629, and outputs the result through a line 622. . The signal output through the line 622 can be defined as a fixed codebook c _G (n) obtained by multiplying the fixed codebook vector of the base layer 600 by the quantization gain value G _c . The quantization gain value G _c is provided from a gain value quantizer 629.

  If an adaptive codebook index is applied through line 613, adaptive codebook 624 outputs an adaptive codebook vector corresponding to the adaptive codebook index. Through line 625, the adaptive codebook vector is provided to a second multiplier 626.

The second multiplier 626, a quantized gain value G _p for gain value of the adaptive codebook, multiplies the adaptive codebook vector transmitted via the line 625, and outputs the result via a line 627. The quantized gain value G _p is provided from a gain value quantizer 629.

  The adder 623 adds the fixed codebook vector input through the line 622 and the adaptive codebook vector input through the line 627 to obtain an excitation signal. The excitation signal is output to the synthesis filter 606 through line 628.

The gain value quantizer 629 quantizes the gain value of the fixed codebook output from the fixed codebook search unit 617 and the gain value of the adaptive codebook output from the pitch analysis unit 612, respectively. Gain value G _c quantized corresponding to the gain value of the fixed codebook is output to the first multiplier 621, a gain value G _p quantized corresponding to the gain value of the adaptive codebook, second multiplier 626 is output. The quantized gain value G _c is also provided to a gain value difference quantizer 643 included in the sound quality enhancement layer 630.

  The sound quality improvement layer 630 is for providing additional bits in addition to the bits provided in the basic layer 600 in order to improve the sound quality restored as in the sound quality improvement layer 130 of FIG. Although FIG. 6 shows a configuration in which one voice enhancement layer 630 is connected to the basic layer 600 for convenience of explanation, a plurality of voice enhancement layers can be connected to the basic layer 600.

  The sound quality improvement layer 630 includes a fixed codebook contribution calculation unit 631, a third adder 633, a synthesis filter 634, a cognitive weighting filter 637, a fixed codebook search unit 639, a fixed codebook 641, and a gain value difference quantizer 643. , And a third multiplier 644.

If the fixed codebook c _G (n) obtained by multiplying the fixed codebook vector by the quantization gain value G _c is received from the first multiplier 621 of the base layer 600, the fixed codebook contribution calculating unit 631 , The fixed codebook contribution y ₂ (n) is calculated by Equation (16).

数式（１６）で、Ｎは副フレームを構成するサンプル数によって決定される。したがって、ピッチ寄与度の除去部６１５で説明したように、副フレームの大きさが４０サンプルである場合にＮは４０である。数式（１６）で、ｈ（ｎ）は合成フィルタのインパルス応答である。固定コードブック寄与度計算部６３１で計算された固定コードブック寄与度は、ライン６３２を通じて第３加算器６３３に提供される。

In Equation (16), N is determined by the number of samples constituting the subframe. Therefore, as described in the pitch contribution removal unit 615, N is 40 when the size of the subframe is 40 samples. In Expression (16), h (n) is an impulse response of the synthesis filter. The fixed codebook contribution calculated by the fixed codebook contribution calculator 631 is provided to the third adder 633 through a line 632.

  The third adder 633 generates the fixed codebook contribution provided through the line 632 and the synthesis filter 634 through the line 635 from the target signal required for the fixed codebook search of the base layer 600 provided through the line 616. A signal obtained by removing the provided composite signal is output.

  When the synthesis filter 634 receives the fixed codebook obtained by multiplying the quantized gain value A of the sound quality enhancement layer 630 by the vector of the fixed codebook through the line 647, the synthesis filter 634 extracts the LPC coefficient and calculates the vector. Using the LPC coefficients extracted and quantized by the quantizer 604, a signal obtained by synthesizing the input fixed codebook signal is output.

  Here, the gain value A is expressed as follows.

認知加重フィルタ６３７は、ライン６３６を通じて入力される信号を、認知加重フィルタ６０１のようにフィルタリングして、音質向上階層６３０で固定コードブック探索のために要求される対象信号を出力する。対象信号は、ライン６３８を通じて固定コードブック探索部６３９に伝送される。

The cognitive weighting filter 637 filters the signal input through the line 636 as the cognitive weighting filter 601 and outputs the target signal required for the fixed codebook search in the sound quality enhancement layer 630. The target signal is transmitted to the fixed codebook search unit 639 through the line 638.

The fixed codebook search unit 639 searches the fixed codebook based on the input target signal like the fixed codebook search unit 617 of the base layer 600 to obtain the fixed codebook index and gain value. The obtained fixed codebook index is transmitted to the fixed codebook 641 while being transmitted to the multiplexer 650 through the line 640. The gain value G _CE of the fixed codebook is transmitted to a gain value difference quantizer 643 through a line 642.

  Fixed codebook 641 outputs a fixed codebook vector of sound quality enhancement layer 630 based on the input fixed codebook index. The fixed codebook vector may include pulse position information m and code information s. The fixed codebook vector output from the fixed codebook 641 is provided to the third multiplier 644. The position of the pulse of the fixed codebook vector output from the fixed codebook 619 of the base layer 600 and the position of the pulse of the fixed codebook vector output from the fixed codebook 641 of the sound quality enhancement layer 630 may be the same.

Quantizer gain difference 643, a gain value G _C a gain value of the fixed codebook output from the gain value quantizer 629 of the base layer 600 by quantizing the fixed codebook searching unit of the speech quality enhancement layer 630 The quantized gain value A of the sound quality enhancement layer 630 is obtained by using the log scale difference value between the non-quantized gain value G _CE of the fixed codebook output from 639 and quantized. A gain value A is output.

  FIG. 7 is a block diagram illustrating a preferred embodiment of the quantizer 643 for gain value difference. The gain value quantizer 643 includes a first log scale converter 702, a second log scale converter 706, fourth and fifth multipliers 708 and 711, and a fourth adder 704.

Fixed codebook gain value G _C quantized provided by the gain value quantizer 629 of the base layer 600, when receiving via the line 701, the first log scale converter 702, a fixed codebook gain value G _The log scale converted fixed codebook gain value corresponding to _c is output through line 703.

If the unquantized gain value G _CE output from the fixed codebook search unit 639 of the sound quality enhancement layer 630 is input through the line 705, the fixed codebook log-scale converted by the second load scale conversion unit 706 The gain value is output through line 707.

  The fourth multiplier 708 multiplies the log scale converted fixed codebook gain value input through the line 707 by the gain value difference adjustment value B and outputs the multiplied result through the line 708.

  Here, the gain value difference adjustment value B is expressed as follows.

第４加算器７０４は、ライン７０３を通じて入力される固定コードブック利得値とライン７０８を通じて入力される固定コードブック利得値間の差値を、ライン７１０を通じて出力する。

The fourth adder 704 outputs a difference value between the fixed codebook gain value input through the line 703 and the fixed codebook gain value input through the line 708 through the line 710.

The fifth multiplier 711 generates the log scale gain value difference G _DIFF 712 by multiplying the input gain value difference by the scale expansion element 10.

  The operation process of the gain value difference quantization 643 described above can be defined as Equation (19).

数式（１９）でＧ_ｃは、利得値量子化器６２９によって量子化された固定コードブックの利得値であり、Ｇ_ＣＥは固定コードブック探索部６３９から出力される量子化されていない利得値である。また、利得値差調整値Ｂは、ログスケール利得値間の差値の動的範囲を最小にする調整値である。利得値差調整値Ｂは、音声符号化器の種類によっていかなる値にもなることができ、実際の例として０.９８７が使われる。

In Equation (19), G _c is the gain value of the fixed codebook quantized by the gain value quantizer 629, and G _CE is the unquantized gain value output from the fixed codebook search unit 639. is there. The gain value difference adjustment value B is an adjustment value that minimizes the dynamic range of the difference value between the log scale gain values. The gain value difference adjustment value B can be any value depending on the type of speech encoder, and 0.987 is used as an actual example.

  Since the log scale gain value difference 712 generated through the calculation of Equation (19) is an analog signal, it is quantized by a 3-bit scalar quantizer. Using the result quantized by the 3-bit scalar quantizer, the quantized fixed codebook gain value A of the sound quality enhancement layer 630 is output. The quantized gain value A is output to the multiplexer 650 through the line 646 while being output to the third multiplier 644 through the line 645.

  The third multiplier 644 multiplies the fixed codebook vector provided from the fixed codebook 641 by the quantized fixed codebook gain value A of the sound quality enhancement layer 630 provided from the gain value difference quantizer 643. Then, the multiplication result is provided to the synthesis filter 634 as a line 647.

  The multiplexer 650 is a fixed codebook of the sound quality improvement layer provided from the LPC coefficient quantization information, the fixed codebook index, the adaptive codebook index, the gain value quantization information and the sound quality improvement layer 630 provided from the base layer 600. The quantization information of the index and gain value difference is multiplexed, and the result is output to the bit stream.

  The bit streams of the basic layer 600 and the sound quality improvement layer 630 are transmitted separately. That is, as shown in FIG. 6, the bit stream of the sound quality enhancement layer 630 is transmitted after the bit stream of the base layer 600. Accordingly, the bit stream can be easily separated at a bit rate required for the decoding apparatus according to network traffic conditions. For example, when the channel characteristics on the decoding device side are poor and only the base layer bit stream can be received, the decoding device transmits the base layer of the bit stream transmitted by the bit rate extended speech encoding device of FIG. Only bitstreams can be received.

  FIG. 8 is a block diagram of a bit rate extended speech decoding apparatus according to another embodiment of the present invention. Referring to FIG. 8, the bit rate extended speech decoding apparatus includes a demultiplexer 802, an LPC coefficient decoding unit 803, a gain value decoding unit 804, a first fixed codebook decoding unit 805, an adaptive codebook decoding. 806, gain value difference decoding unit 807, second fixed codebook decoding unit 808, multipliers 809, 810, 813, adders 811, 814, selection switch 812, synthesis filter 815, and post-processing unit 816 including.

  The bit rate extended speech decoding apparatus can selectively receive a bit stream transmitted from the bit rate extended speech encoding apparatus. That is, if only the bitstream for the base layer is received, the sound quality of the base layer can be restored, and if both the bitstreams for the base layer and the sound quality improvement layer are received, further improved sound quality can be provided.

  The demultiplexer 802 demultiplexes the received bit stream 801 into information of each component and outputs the information. That is, the demultiplexer 802 decodes LPC coefficient quantization information to the LPC coefficient decoding unit 803, gain value quantization information to the gain value decoding unit 804, and gain value difference quantization information to decode the gain value difference. In the second fixed codebook decoding unit 808 for the fixed codebook index of the sound quality enhancement layer 630 and the adaptive codebook index in the first fixed codebook decoding unit 805 for the fixed codebook index of the basic layer 600. Are provided to the adaptive codebook decoding unit 806, respectively.

  The structure of the LPC coefficient decoding unit 803 is determined by the LPC coefficient extraction and vector quantizer 604 on the encoder side, and restores the LPC coefficient from the input LPC coefficient quantization information. The restored LPC coefficient is provided to the synthesis filter 815 and the post-processing unit 816.

The structure of gain value decoding section 804 is determined by gain value quantizer 629 on the encoding device side. The gain value decoding unit 804 decodes the input gain value quantization information. The gain value quantization information includes an adaptive codebook gain value and a fixed codebook gain value. Thus, the adaptive codebook gain value G _P of the basic hierarchical 600 and fixed codebook gain value G _C from the gain value decoding unit 804 are output.

  The first fixed codebook decoding unit 805 decodes the input first fixed codebook index and outputs the first fixed codebook. The fixed codebook decoding method is determined by the search method in fixed codebook search unit 617 of the encoding device.

  The adaptive codebook decoding unit 806 decodes the input adaptive codebook index and outputs an adaptive codebook.

  The LPC coefficient decoding unit 803, gain value decoding unit 804, fixed codebook decoding unit 805, and adaptive codebook decoding unit 806 described above are encoded in the base layer 600 transmitted from the demultiplexer 802. It may be defined as a decoding unit that decodes information.

  The operations of the gain value difference decoding unit 807 and the second fixed codebook decoding unit 808 depend on the network traffic state and the processing capacity of the receiving terminal.

  If it is determined that the gain value difference decoding unit 807 and the second fixed codebook decoding unit 808 operate, the gain value difference decoding unit 807 determines the gain value difference quantization information to be input. Decode. The second fixed codebook decoding unit 808 decodes the input second fixed codebook index. The gain value difference decoding method is determined by a gain value difference quantizer 643 on the encoding device side.

  The decoding method performed by the second fixed codebook decoding unit 808 is determined by the second fixed codebook search unit 631 on the encoding device side. The gain value difference decoding unit 807 and the second fixed codebook decoding unit 808 may be defined as decoding units that decode the encoded information in the sound quality enhancement layer 630 transmitted from the demultiplexer 902.

The multiplier 809 multiplies the fixed codebook gain value G _c of the base layer 600 restored by the gain value decoding unit 804 to the fixed codebook of the first fixed codebook decoding unit 805 output the base layer by Output base layer fixed codebook vector.

  The multiplier 810 uses the fixed codebook gain value A in the sound quality enhancement layer 630 restored by the gain value difference decoding unit 807 as the fixed code of the sound quality improvement layer output from the second fixed codebook decoding unit 808. Multiplies the book and outputs a fixed codebook vector of the sound quality enhancement layer.

  The adder 811 adds the fixed codebook vector of the basic layer output from the multiplier 809 and the fixed codebook vector of the sound quality improvement layer output from the multiplier 810. Accordingly, the fixed codebook pulse in the decoding apparatus has an algebraic codebook pulse structure having a multiple size by accumulating the algebraic codebooks of the basic layer and the sound quality enhancement layer. Accumulating the algebraic codebook is intended to compensate for the disadvantages that occur with existing fixed codebook structures in which every pulse of the fixed codebook has the same magnitude.

  The selection switch 812 selectively transmits the signal output from the adder 811 and the base layer fixed codebook vector output from the multiplier 809. That is, when the decoding apparatus does not operate in the sound quality enhancement layer, the selection switch 812 selects and transmits the fixed codebook vector of the basic layer output from the multiplier 809. When the encoding apparatus operates in the sound quality improvement layer, the selection switch 812 selects and transmits the signal output from the adder 811.

Multiplier 813, an adaptive codebook vector is multiplied by a gain value G _p of the adaptive codebook output from the gain value decoding unit 804 to the decoded adaptive codebook output from the adaptive codebook decoding unit 806 Output.

  The adder 814 adds the fixed codebook vector selected by the selection switch 812 and the adaptive codebook vector output from the multiplier 813 to generate a restored excitation signal.

  The multiplier 810, the adder 811 and the selection switch 812 described above include a unit for decoding the encoding information of the base layer and a unit for decoding the encoding information of the sound quality improvement layer according to the operating environment of the decoding device. Can be defined as operation units for calculating the decoded signals.

  The synthesis filter 815 synthesizes the excitation signal provided from the adder 814 using the restored LPC coefficient provided from the LPC coefficient decoding unit 803 to restore the speech signal.

  The post-processing unit 816 restores the audio signal transmitted from the synthesis filter 815. That is, the post-processing unit 816 uses a high-pass filter for filtering the signal output from the synthesis filter 815 using the LPC coefficient provided from the LPC coefficient decoding unit 803 to restore the audio signal. To do.

  The synthesis filter 815 and the post-processing unit 816 described above may be defined as a restoration unit that synthesizes the signal output from the arithmetic unit with the LPC coefficient output from the LPC coefficient decoding unit 803 to restore the speech signal.

  FIG. 9 shows the positions of pulses searched by the fixed codebook search 901 in the basic layer and the pulse positions searched by the fixed codebook search 905 in the sound quality improvement layer in the speech signal encoding apparatus of FIG. FIG. 9 is a diagram for explaining the magnitude of a pulse restored by the speech signal decoding apparatus of FIG. 8 using a fixed codebook vector based thereon.

Referring to FIG. 9, the fixed codebook vector 902 provided in the first fixed codebook decoding unit 805, the fixed codebook gain value G _c, which is provided by the gain value decoding unit 804, the multiplier 809 A base layer fixed codebook vector 904 multiplied by the gain value is generated by multiplication.

A fixed codebook vector 906 provided in the second fixed codebook decoding unit 808, a gain value G _CE provided by the decoding unit 807 of the gain difference is multiplied by a multiplier 810, a gain value is multiplied A sound quality improvement hierarchical fixed codebook vector 908 is generated. The adder 811 generates a fixed codebook vector 910 obtained by adding the sound quality improvement hierarchical fixed codebook vector 908 and the basic hierarchical fixed codebook vector 904.

  As shown in the pulse structure of FIG. 9, the fixed codebook vector 904 of the basic layer and the fixed codebook vector 908 of the sound quality enhancement layer are input to the adder 811 to be the final fixed codebook of the sound quality enhancement layer. 910 is generated. Since the final sound quality enhancement hierarchical fixed codebook 910 is obtained by adding two fixed codebook vectors having different gain values, a fixed codebook having multiple sizes can be formed, and better sound quality can be provided.

  FIG. 10 is a flowchart of a bit rate extended speech encoding method according to another embodiment of the present invention.

  In operation 1001, the audio signal encoding apparatus preprocesses an input audio signal as in the preprocessing unit 602 of FIG. In operation 1002, the speech signal encoding apparatus extracts LPC coefficients from the preprocessed speech signal and generates quantization information of the extracted LPC coefficients.

  In operation 1003, the speech signal encoding apparatus detects a residual signal of an LPC coefficient from the preprocessed signal through a synthesis filter 606. In operation 1004, the speech signal encoding apparatus filters the detected residual signal as in the perceptual weighting filter 610 of FIG. 6 and outputs a perceptual weighted signal.

  In operation 1005, the speech signal encoding apparatus analyzes the pitch of the cognitive weighted signal as in the pitch analysis unit 612 of FIG. 6 and the pitch contribution removal unit 615 of FIG. 6. The result is used to remove the pitch contribution from the cognitive weighted signal to generate an adaptive codebook gain value and an adaptive codebook index.

  In operation 1006, the speech signal encoding apparatus searches the base layer fixed codebook to search for a fixed codebook gain value and a fixed codebook index as in the fixed codebook search unit 617 of the base layer 600 of FIG. And generate

  In operation 1007, the speech signal encoding apparatus quantizes the detected fixed codebook gain value and the detected adaptive codebook gain value as in the gain value quantizer 629 of FIG. To do.

  In operation 1008, the speech signal encoding apparatus uses the vector quantized LPC coefficients to generate the excitation signal of the fixed codebook vector and the adaptive codebook vector generated in the base layer 600, as shown in FIG. Composite as in filter 606.

  In operation 1009, the speech signal encoding apparatus removes the influence of the target signal for the fixed codebook search in the base layer 600 and the previous LPC synthesized signal in the sound quality improvement layer 630. A target signal for fixed codebook search as in the codebook search unit 639 is generated. That is, from the target signal detected in the base layer 600, the fixed codebook contribution degree in the base layer and the previous LPC synthesized signal detected in the sound quality improvement layer 630 are removed to obtain the target signal in the sound quality improvement layer.

  In operation 1010, the speech signal encoding apparatus performs a fixed codebook search in the sound quality improvement layer 630 using the target signal detected in operation 1009, and the fixed codebook gain value and sound quality in the sound quality improvement layer. A fixed codebook index of the improvement hierarchy is generated respectively.

  In operation 1011, the speech signal encoding apparatus quantizes a log scale difference between a quantized fixed codebook gain value of a base layer and an unquantized fixed codebook gain value of a sound quality enhancement layer. . The fixed codebook search and gain value quantization process in the sound quality enhancement layer described above may be performed in a plurality of layers by providing a plurality of sound quality enhancement layers. If the sound quality improvement layer processing is performed in a plurality of layers, the quality of the restored audio signal can be improved.

  In operation 1012, the speech signal encoding apparatus passes the fixed codebook vector (or excitation signal) generated in the sound quality enhancement layer to the synthesis filter 634 in FIG. 6 and outputs a synthesized signal.

  In operation 1013, the speech signal encoding apparatus performs linear prediction coefficient quantization information obtained through the above-described steps, base layer fixed codebook index, base layer adaptive codebook index, base layer fixed codebook gain. Value, base layer adaptive codebook gain value, sound quality enhancement layer fixed codebook index, and gain value difference quantization information are multiplexed to obtain a bitstream, and the bitstream is sent to the audio signal decoding device side To do.

  FIG. 11 is a flowchart of a bit rate enhanced speech decoding method according to another embodiment of the present invention.

  In operation 1101, the audio signal decoding apparatus demultiplexes the received bitstream into information of each component as in the demultiplexer 802 of FIG.

  In operation 1102, the audio signal decoding apparatus decodes the demultiplexed signal. That is, the LPC coefficient decoding unit 803, gain value decoding unit 804, first fixed codebook decoding unit 805, adaptive codebook decoding unit 806, gain value difference decoding unit 807, and second fixed code in FIG. Like the book decoding unit 808, the demultiplexed signal is decoded.

  In operation 1103, the audio signal decoding apparatus selectively transmits a fixed codebook of a sound quality enhancement layer or a fixed codebook of a basic layer according to the operating conditions of the audio signal decoding apparatus, and also selectively transmits a gain value. Is done. That is, if the speech signal decoding apparatus operates in the sound quality enhancement layer, the base layer is divided into the fixed codebook of the sound quality improvement layer multiplied by the gain value of the fixed codebook of the restored sound quality improvement layer and the fixed codebook of the base layer. The fixed codebook multiplied by the gain value of the fixed codebook is added, and the result is transmitted. On the other hand, if the speech signal encoding device does not operate in the sound quality enhancement layer, the speech signal decoding device obtains the decoded base layer fixed codebook by the gain value of the base layer fixed codebook. The fixed codebook is transmitted.

  In operation 1104, the speech signal decoding apparatus synthesizes the fixed codebook selectively transmitted in operation 1103 using the LPC coefficients decoded in operation 1102.

  In operation 1105, the speech signal decoding apparatus generates a speech signal restored by post-processing as in the post-processing unit 816.

  The present invention has been described above with a focus on preferred embodiments thereof. Those skilled in the art to which the present invention pertains can understand that the present invention can be embodied in a modified form without departing from the essential characteristics of the present invention. Accordingly, the disclosed embodiments are to be considered in an illustrative, not a limiting sense. The scope of the present invention is described in the scope of claims rather than the above description, and all differences within the equivalent scope should be construed as being included in the present invention.

  The present invention is applicable to a speech encoding / decoding device that uses a CELP algorithm and has a basic layer and a sound quality improvement layer.

1 is a block diagram of a bit rate extended speech encoding apparatus according to an embodiment of the present invention. FIG. 2 is an exemplary diagram of a pulse position searched by a basic layer fixed codebook search unit shown in FIG. 1 and a pulse position searched by a sound quality improvement layer fixed codebook search unit. 1 is a block diagram of a bit rate extended speech decoding apparatus according to an embodiment of the present invention. 3 is a flowchart of a bit rate extended speech encoding method according to an embodiment of the present invention. 3 is a flowchart of a bit rate extended speech decoding method according to an embodiment of the present invention. FIG. 5 is a block diagram of a bit rate extended speech encoding apparatus according to another embodiment of the present invention. FIG. 7 is a block diagram of a quantizer for gain value difference of a sound quality improvement layer in the bit rate extended speech encoding apparatus of FIG. 6. FIG. 6 is a block diagram of a bit rate extended speech decoding apparatus according to another embodiment of the present invention. FIG. 9 is an exemplary diagram of a pulse position searched by a base layer fixed codebook search and a pulse position searched by a sound quality improvement layer fixed codebook search in the bit rate extended speech decoding apparatus of FIG. 8. 7 is a flowchart of a bit rate extended speech encoding method according to another embodiment of the present invention. 5 is a flowchart of a bit rate enhanced speech decoding method according to another embodiment of the present invention.

Explanation of symbols

100 Base layer 101, 103, 105, 107, 109, 111, 113, 114, 116, 118, 118 ', 118'', 120, 122, 125, 127, 135 Line 102 Preprocessing unit 104 LPC coefficient extraction and Vector quantizer 106 Synthesis filter 108 Subtractor 110 Cognitive weighting filter 112 Pitch analysis unit 115 Pitch contribution removal unit 117 Fixed codebook search unit 119 Fixed codebook 121 First multiplier 123 Adder 124 Adaptive codebook 126 Second Multiplier 129 Gain value quantization unit 130 Sound quality enhancement layer 131 Fixed codebook search unit 132 Gain value 134 Gain value difference quantizer 140 Multiplexer _{Gc, Gp} Quantized gain value

Claims (19)

In an audio signal encoding device,
Filtering the input speech signal using linear predictive coding and generating an excitation signal of the filtered speech signal by fixed codebook search and adaptive codebook search;
Defined by the correlation d (n) between the target signal and the impulse response signal required for the fixed codebook search detected in the basic layer, the code of the pulse and the correlation d (n), Each pulse, each based on a parameter obtained by a fixed codebook search in the basic layer, including a correlation C corresponding to the size information of the correlation d (n) and an energy E of the impulse response signal A sound quality enhancement layer for searching a fixed codebook by searching an algebraic codebook in which a pulse code and a position of each pulse are associated with each other; and
A speech signal encoding apparatus including a multiplexer that multiplexes a signal generated in the base layer and a signal generated in the sound quality improvement layer and outputs the multiplexed signal.   The speech signal encoding apparatus according to claim 1, wherein the fixed codebook search in the basic layer and the fixed codebook search in the sound quality improvement layer are performed using an algebraic codebook.   The sound quality enhancement layer quantizes a difference between a first gain value obtained by a fixed codebook search performed by the basic layer and a second gain value obtained by a fixed codebook search in the sound quality enhancement layer. The speech signal encoding apparatus according to claim 1, further comprising a function.   The multiplexer includes quantization information of linear predictive coding coefficients generated in the base layer, a fixed codebook index in the base layer, an adaptive codebook index in the base layer, and a fixed codebook in the base layer. Quantization information of gain values and quantization information of adaptive codebook gain values in the base layer, fixed codebook index generated in the sound quality enhancement layer, and fixed codebook gain values and sound quality enhancement layer in the base layer The apparatus according to claim 1, wherein the quantized information for the difference value between the fixed codebook gain values is multiplexed. If there are a plurality of sound quality enhancement layers, the multiplexer multiplexes quantized information related to a difference value between a fixed codebook index output from a plurality of sound quality enhancement layers and the fixed codebook gain value. The audio signal encoding apparatus according to claim 4 , wherein the apparatus is an audio signal encoding apparatus. In an audio signal encoding device,
A base layer for performing linear predictive coding filtering on an input speech signal and generating an excitation signal corresponding to the filtered speech signal by fixed codebook search and adaptive codebook search;
Defined by the correlation d (n) between the target signal and the impulse response signal required for the fixed codebook search detected in the basic layer, the code of the pulse and the correlation d (n), Each pulse based on a parameter generated by a fixed codebook search in the base layer, including a correlation C corresponding to size information of the correlation d (n) and an energy E of an impulse response signal . A fixed codebook search unit that searches for a fixed codebook by searching an algebraic codebook that associates the code of each pulse and the position of each pulse ;
A difference between a first fixed codebook gain value generated by the fixed codebook search of the base layer and a second fixed codebook gain value output from the fixed codebook search unit is detected, and the detected difference is detected. A plurality of sound quality enhancement layers including a quantizer for gain value difference to be quantized,
An audio signal encoding apparatus including a multiplexer that multiplexes a signal generated in the base layer and a signal generated in the sound quality improvement layer. In the audio signal decoding device for decoding an audio signal encoded by the encoding device according to any one of claims 1 to 6 , divided into a basic layer and at least one sound quality improvement layer,
A first decoding unit for decoding encoded information in a base layer of the encoded audio signal;
A second decoding unit for restoring encoded information in a sound quality enhancement layer of the encoded audio signal according to an operating environment of the audio signal decoding device;
An arithmetic unit for calculating a signal restored by the first decoding unit and a signal restored by the second decoding unit according to an operating environment of the audio signal decoding device;
An audio signal decoding apparatus comprising: an audio signal restoration unit that synthesizes a signal output from the arithmetic unit using the linear predictive coding coefficient output from the first decoding unit to restore an audio signal. The first decoding unit is
A linear predictive coding coefficient composite unit for decoding quantization information of the linear predictive coding coefficient included in the coding information in the base layer;
A first fixed codebook decoding unit for decoding a fixed codebook index included in the encoding information in the base layer;
An adaptive codebook decoding unit for decoding an adaptive codebook index included in the encoding information in the base layer;
The speech signal decoding apparatus according to claim 7 , further comprising: a gain value decoding unit that decodes a fixed codebook gain value and an adaptive codebook gain value included in the encoding information in the base layer. The second decoding unit is
A gain value difference decoding unit for decoding quantization information of a difference between fixed codebook gain values included in encoding information in the speech enhancement layer;
The audio signal decoding apparatus according to claim 8 , further comprising: a second fixed codebook decoding unit that decodes a fixed codebook index included in the encoded information in the sound quality enhancement layer. The arithmetic unit is
A first adder for adding a difference between the decoded fixed codebook gain value output from the gain value decoding unit and the decoded gain value output from the decoding unit of the gain value difference;
A first selection switch for transmitting the decoded fixed codebook gain value output from the gain value decoding unit or the gain value output from the first adder according to an operating condition of the audio signal decoding device;
The decoded fixed codebook of the sound quality enhancement layer output from the second fixed codebook decoding unit and the decoded basic layer fixed codebook output from the first fixed codebook decoding unit are added. A second adder;
A second selection switch for transmitting the signal output from the second adder or the decoded fixed codebook output from the first fixed codebook decoding unit according to the operating condition of the audio signal decoding device;
A first multiplier for multiplying a gain value of the decoded adaptive codebook output from the adaptive codebook decoding unit and the decoded adaptive codebook output from the gain value decoding unit;
A second multiplier that multiplies the signal output from the first selection switch and the signal output from the second selection switch;
The speech signal decoding apparatus according to claim 9 , further comprising: a third adder that adds a signal output from the first multiplication branch and a signal output from the second multiplication branch. The audio signal restoration unit includes:
A synthesis filter that synthesizes a signal output from the third adder using the linear predictive coding coefficient;
The speech signal decoding apparatus according to claim 10 , further comprising: a post-processing unit for obtaining the restored speech signal using the linear predictive coding coefficient and a signal output from the synthesis filter. The second decoding unit is
A gain value difference decoding unit for decoding quantization information of a difference between fixed codebook gain values included in encoding information in the speech enhancement layer;
The audio signal decoding device according to claim 7 , further comprising: a fixed codebook decoding unit that decodes a fixed codebook index included in the encoding information in the sound quality enhancement layer. The audio signal encoding method is
Extracting a linear predictive coding coefficient of a speech signal input in a basic layer, and generating an excitation signal corresponding to the input speech signal by a fixed codebook search and an adaptive codebook search;
In at least one sound quality enhancement layer, a correlation d (n) between the target signal and impulse response signal required for the fixed codebook search detected in the basic layer, a pulse code, and the correlation d ( n), the correlation degree C corresponding to the size information of the correlation degree d (n), and the energy E of the impulse response signal, and generated by searching the fixed codebook in the base layer Searching for a fixed codebook by searching an algebraic codebook that associates each pulse, the sign of each pulse and the position of each pulse based on a parametric variable;
Multiplexing a signal generated by the base layer and the sound quality enhancement layer. The method of claim 13 , wherein the sound quality enhancement hierarchical processing step is performed in a plurality of steps. The sound quality enhancement layer processing step quantizes a difference value between the fixed codebook gain value obtained by the fixed codebook search in the basic layer processing step and the gain value obtained by the fixed codebook search of the sound quality improvement layer. The audio signal encoding method according to claim 13 , wherein: The parametric variable includes a first correlation between the target signal and the impulse response generated in the base layer processing step, a second correlation corresponding to the magnitude of the first correlation, and an energy of the impulse response signal. The audio signal encoding method according to claim 13 , wherein: An audio signal decoding method for decoding an audio signal encoded into a base layer and at least one sound quality improvement layer by the audio signal encoding method according to any one of claims 13 to 16 ,
Decoding the encoded audio signal;
Selectively transmitting the codebook for the base layer and the codebook for the sound quality enhancement layer decoded in the decoding step according to the operation condition of the speech signal encoding;
Generating a restored speech signal by synthesizing the selectively transmitted codebook and the linear prediction coefficient decoded in the decoding step. The decoding step includes
Claim 17, wherein the encoded audio signal demultiplexed into the encoded information related to the coding information and speech quality enhancement layer relating to the base layer, to decode the encoded information demultiplexing Audio signal decoding method. The audio signal decoding method includes:
The difference between the gain value of the fixed codebook in the base layer decoded in the decoding step and the gain value of the fixed codebook included in the decoded sound quality enhancement layer is added to fix the fixed codebook in the sound quality enhancement layer The audio signal decoding method according to claim 18 , further comprising: restoring a gain value of the audio signal.

Images